基于Gabor滤波与区域生长的细胞分割

Cell Segmentation Based on Gabor Filtering and Regional Growth

贴壁细胞图像中细胞大小形状各不相同,细胞内部分区域与细胞边缘具有相近的灰度,部分细胞边缘较细或者断裂,为后续正确的分割计数带来了困难。传统的基于区域的图像分割很容易造成过分割和欠分割,前者得不到完整的细胞边缘,后者使细胞内部出现大量杂质点。为解决以上问题,首先利用Gabor滤波器的方向性滤波特性对细胞边缘进行增强,然后选择细胞边缘对应的高灰度点作为种子点进行8连通约束的区域生长,最后对区域生长后的图像进行形态学闭运算消除小的空洞和毛刺,得到完整的细胞边缘图像。与阈值法和边缘检测法的比较结果表明,该算法分割效果较好且对噪声不敏感。

The wall-stuck cells are different in sizes and shapes, the gray grade of cell inner regions is close to that of cell edges, the parts of the cell edges are thin and broken, leading to the difficulty for afterwards correctly segmenting and counting the number of cells. The traditional image segmentation based on regions was easy to be over segmented or inadequately segmented. The former cannot obtain the complete edge of cell and the latter obtains a number of ira-purity spots. In order to solve the above problem,first,Gabor filter was used to enhance the edges in different direc- tions;then, the high gray grade points corresponding to cell edges were used as seed points to carry out the regional growth with eight connective restrictions;finally,the morphologically closed operation was done for the image after the regional growth to eliminate small cavities and burrs and to get the image with complete cell edge. The results show that this algorithm is of good effect of segmentation and not sensitive to noise compared with threshold method and edge detection method.